Spark SQL SaveMode.Overwrite, getting java.io.File

For spark sql, how should we fetch data from one folder in HDFS, do some modifications, and save the updated data to the same folder in HDFS via Overwrite save mode without getting FileNotFoundException?

import org.apache.spark.sql.{SparkSession,SaveMode}
import org.apache.spark.SparkConf

val sparkConf: SparkConf = new SparkConf()
val sparkSession = SparkSession.builder.config(sparkConf).getOrCreate()
val df = sparkSession.read.parquet("hdfs://xxx.xxx.xxx.xxx:xx/test/d=2017-03-20")
val newDF = df.select("a","b","c")

newDF.write.mode(SaveMode.Overwrite)
     .parquet("hdfs://xxx.xxx.xxx.xxx:xx/test/d=2017-03-20") // doesn't work
newDF.write.mode(SaveMode.Overwrite)
     .parquet("hdfs://xxx.xxx.xxx.xxx:xx/test/d=2017-03-21") // works

FileNotFoundException happens when we read data from the hdfs dir "d=2017-03-20", and save (SaveMode.Overwrite) updated data to the same hdfs dir "d=2017-03-20"

Caused by: org.apache.spark.SparkException: Task failed while writing rows
  at org.apache.spark.sql.execution.datasources.FileFormatWriter$.org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask(FileFormatWriter.scala:204)
  at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1$$anonfun$3.apply(FileFormatWriter.scala:129)
  at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1$$anonfun$3.apply(FileFormatWriter.scala:128)
  at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87)
  at org.apache.spark.scheduler.Task.run(Task.scala:99)
  at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:282)
  at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
  at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
  at java.lang.Thread.run(Thread.java:745)
Caused by: java.io.FileNotFoundException: File does not exist: hdfs://xxx.xxx.xxx.xxx:xx/test/d=2017-03-20/part-05020-35ea100f-829e-43d9-9003061-1788904de770.snappy.parquet
It is possible the underlying files have been updated. You can explicitly invalidate the cache in Spark by running 'REFRESH TABLE tableName' command in SQL or by recreating the Dataset/DataFrame involved.
  at org.apache.spark.sql.execution.datasources.FileScanRDD$$anon$1.nextIterator(FileScanRDD.scala:157)
  at org.apache.spark.sql.execution.datasources.FileScanRDD$$anon$1.hasNext(FileScanRDD.scala:102)
  at org.apache.spark.sql.catalyst.expressions.GeneratedClass$GeneratedIterator.scan_nextBatch$(Unknown Source)
  at org.apache.spark.sql.catalyst.expressions.GeneratedClass$GeneratedIterator.processNext(Unknown Source)
  at org.apache.spark.sql.execution.BufferedRowIterator.hasNext(BufferedRowIterator.java:43)
  at org.apache.spark.sql.execution.WholeStageCodegenExec$$anonfun$8$$anon$1.hasNext(WholeStageCodegenExec.scala:377)
  at org.apache.spark.sql.execution.datasources.FileFormatWriter$SingleDirectoryWriteTask.execute(FileFormatWriter.scala:243)
  at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$3.apply(FileFormatWriter.scala:190)
  at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$3.apply(FileFormatWriter.scala:188)
  at org.apache.spark.util.Utils$.tryWithSafeFinallyAndFailureCallbacks(Utils.scala:1341)
  at org.apache.spark.sql.execution.datasources.FileFormatWriter$.org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask(FileFormatWriter.scala:193)
  ... 8 more

The following tries still get the same error, how should I solve this problem by using spark sql? Thank you!

val hdfsDirPath = "hdfs://xxx.xxx.xxx.xxx:xx/test/d=2017-03-20"

val df= sparkSession.read.parquet(hdfsDirPath)

val newdf = df
newdf.write.mode(SaveMode.Overwrite).parquet(hdfsDirPath)

val df= sparkSession.read.parquet(hdfsDirPath)
df.createOrReplaceTempView("orgtable")
sparkSession.sql("SELECT * from orgtable").createOrReplaceTempView("tmptable")

sparkSession.sql("TRUNCATE TABLE orgtable")
sparkSession.sql("INSERT INTO orgtable SELECT * FROM tmptable")

val newdf = sparkSession.sql("SELECT * FROM orgtable")
newdf.write.mode(SaveMode.Overwrite).parquet(hdfsDirPath)

val df= sparkSession.read.parquet(hdfsDirPath)
df.createOrReplaceTempView("orgtable")
sparkSession.sql("SELECT * from orgtable").createOrReplaceTempView("tmptable")

sparkSession.sql("REFRESH TABLE orgtable")
sparkSession.sql("ALTER VIEW tmptable RENAME TO orgtable")

val newdf = sparkSession.sql("SELECT * FROM orgtable")
newdf.write.mode(SaveMode.Overwrite).parquet(hdfsDirPath)

回答1:

I solved this , first I write my Dataframe to a temp directory , and delete the source I reading , and rename the temp directory to source name . QAQ

回答2:

Why don't you just cache it after reading it. Saving it to another file directory and then moving the directory might entail some extra permissions. I also have been forcing an action as well, like a show().

val myDF = spark.read.format("csv")
    .option("header", "false")
    .option("delimiter", ",")
    .load("/directory/tofile/")


myDF.cache()
myDF.show(2)

回答3:

val dfOut = df.filter(r => r.getAs[Long]("dsctimestamp") > (System.currentTimeMillis() - 1800000))

In the above line of code, df had an underlying Hadoop partition. Once I had made this transformation (i.e., to dfOut), I could not find a way to delete, rename, or overwrite the underlying partition until dfOut had been garbage collected.

My solution was to keep the old partition, create a new partition for dfOut, flag the new partition as current and then delete the old partition some given time later, after dfOut had been garbage collected.

May not be an ideal solution. I would love to learn a less tortuous way of addressing this issue. But it works.